OpenAI drops 700+ AI math preprints; mathematicians balk

OpenAI posted more than 700 AI-generated mathematical preprints to GitHub on 6 October, aimed at hundreds of high-profile open problems. The output came from an unreleased OpenAI model. Nature reports that the drop split the mathematics community almost immediately.

Michael Harris, a number theorist at Columbia University, put it flatly: "I can't bring myself to spend my time studying the outcome of this process," he told Nature. "So I will not be reading these papers."

Others chose different registers. Alvaro Lozano-Robledo, a number theorist at the University of Connecticut, called the simultaneous volume "staggering" and asked, "Why is OpenAI trying to solve hundreds of math problems to be released all at once?" Ursula Martin, at the University of Oxford, said the company was "tossing the community a messy first draft and expecting us to clean it up." One mathematician reached for a sharper word and called the moment a "slopocalypse."

Not everyone recoiled. Levent Alpöge, a mathematician at Anthropic, acknowledged there would be "sad stories" for researchers whose projects got pre-empted, but told Nature it was "obviously the most significant moment in mathematical history."

OpenAI's stated position, as reported, is that its goal is to provide tools for mathematicians and scientists. The release did not include solutions to the five remaining Millennium Prize Problems. Three AI researchers we follow circulated the Nature piece the week it ran.

The article was written by Davide Castelvecchi and updated twice after publication.